A simulation is the imitation of the operation of a real-world process or system over time.
Simulations require the use of models; the model represents the key characteristics or behaviors of the selected system or process, whereas the simulation represents the evolution of the model over time.
The AFRS Call Centre has been working nonstop since it’s opening at the beginning of the pandemic.
Since the beginning there has been a steady increase of calls per month, and we expect that trend to continue. Reference Liz Moakes Demand and Funding Position Statement for specifics.
Even through this increase the Call Centre has performed consistently overtime in managing this call overload.
This simulation will look at the effect of increasing the total number of operators in the Call Centre for an entire day, simulated over multiple days.
First our model picks a number from a poisson distribution based off the hourly distributions. A poisson distribution tracks the expected number of times something will help in a given amount of time. (i.e. how many calls will a call centre receive in an hour)
Next the model uses that number to generate n random numbers in the time span of that hour, so if the number generated was 4, then four random times between midnight and 1 a.m. would be generated, they could be close together or far apart. This simulates a semi-realistic call centre, since it assumes that each person calling is independent of another person calling.
DES Works as a pipeline, the same way an interaction in the call centre might occur.
A caller calls, and is met with either a wait time or an operator. A strict rule of 60 seconds was set for callers to abandon. (This is stricter than realistic)
If an operator is available they will answer the call, the length of the call is determined from a distribution of the average of the average monthly call lengths plus an additional wrap-up time.
After that length of time has passed the operator will once again become available for another call.
This simulation runs for 133 days with n operators per day
Considers professional referrals, which then are followed-up after two hours. These follow-ups take priority over incoming calls.
Does not include downtime for supervision, etc. so results are UNDERESTIMATES
We can see that overtime we have operators being used, this is extremely volatile, however we can see spikes around the early hours of the morning as well as a rapid increase as we reach evening and a slow decrease at the end of the day.
The utilisation is around 4 operators consistently on the phone, however this is the AVERAGE of the whole day. If you think about it, this makes sense. 4-5 operators consistently on the phone, 1 operator on a stress/mental health break, and 2-3 operators completing supervision, preparing to follow-up on a referral, etc.
Even though utilisation for the service is lower than what some might consider acceptable, the reality is that the operators in the call centre cannot be viewed in the same light as another resource such as beds in a ward.
Operators are there to triage high-risk patients, a missed call has a high probability to lead to personal harm, harm to others or even death. Therefore, it is much better for an operator to be waiting and available, than all operators being utilised an not having someone available to answer a call.
Due to the nature of the calls, these operators require extra time for stress breaks, mental health breaks, etc.
Due to these reasonings, a utilisation average of 45-55% for the day is a more than reasonable compromise.
The distribution of the number of calls per hour shows a high level of similarity between both our simulated data and the real data.
| Abandoned | count | Percent_of_Total |
|---|---|---|
| FALSE | 15607 | 0.95 |
| TRUE | 899 | 0.05 |
As we can see one of our most important KPIs, Calls Abandoned after 60s, is down to 5%. This is an even stricter cutoff since our simulation literally forced callers to leave if they were waiting for a minute. There is potential that our results could be even better.
The utilization is between 45-55%, this is to account for one the stress breaks that might befall call operators dealing with mental health patients, supervision, and the downtime that is the reality of a call centre, especially in the early hours of the morning. Remember this is the AVERAGE utilization for ALL hours.
“This simulation doesn’t reflect reality”, well in reality, it does. It was built off of the real distribution of the data, so although it has not actually happened, this is what is MOST LIKELY to happen.
If there any questions please feel free to contact:
Hansel Palencia - Senior Information Analyst - Informatics